mirror of
https://github.com/msberends/AMR.git
synced 2024-12-25 20:46:11 +01:00
226 lines
9.7 KiB
R
226 lines
9.7 KiB
R
# ==================================================================== #
|
|
# TITLE #
|
|
# Antimicrobial Resistance (AMR) Analysis #
|
|
# #
|
|
# SOURCE #
|
|
# https://github.com/msberends/AMR #
|
|
# #
|
|
# LICENCE #
|
|
# (c) 2018-2020 Berends MS, Luz CF et al. #
|
|
# #
|
|
# This R package is free software; you can freely use and distribute #
|
|
# it for both personal and commercial purposes under the terms of the #
|
|
# GNU General Public License version 2.0 (GNU GPL-2), as published by #
|
|
# the Free Software Foundation. #
|
|
# #
|
|
# We created this package for both routine data analysis and academic #
|
|
# research and it was publicly released in the hope that it will be #
|
|
# useful, but it comes WITHOUT ANY WARRANTY OR LIABILITY. #
|
|
# Visit our website for more info: https://msberends.github.io/AMR. #
|
|
# ==================================================================== #
|
|
|
|
#' Determine bug-drug combinations
|
|
#'
|
|
#' Determine antimicrobial resistance (AMR) of all bug-drug combinations in your data set where at least 30 (default) isolates are available per species. Use [format()] on the result to prettify it to a publicable/printable format, see Examples.
|
|
#' @inheritSection lifecycle Stable lifecycle
|
|
#' @inheritParams eucast_rules
|
|
#' @param combine_IR logical to indicate whether values R and I should be summed
|
|
#' @param add_ab_group logical to indicate where the group of the antimicrobials must be included as a first column
|
|
#' @param remove_intrinsic_resistant logical to indicate that rows with 100% resistance for all tested antimicrobials must be removed from the table
|
|
#' @param FUN the function to call on the `mo` column to transform the microorganism IDs, defaults to [mo_shortname()]
|
|
#' @param translate_ab a character of length 1 containing column names of the [antibiotics] data set
|
|
#' @param ... arguments passed on to `FUN`
|
|
#' @inheritParams rsi_df
|
|
#' @inheritParams base::formatC
|
|
#' @details The function [format()] calculates the resistance per bug-drug combination. Use `combine_IR = FALSE` (default) to test R vs. S+I and `combine_IR = TRUE` to test R+I vs. S.
|
|
#' @export
|
|
#' @rdname bug_drug_combinations
|
|
#' @return The function [bug_drug_combinations()] returns a [`data.frame`] with columns "mo", "ab", "S", "I", "R" and "total".
|
|
#' @source \strong{M39 Analysis and Presentation of Cumulative Antimicrobial Susceptibility Test Data, 4th Edition}, 2014, *Clinical and Laboratory Standards Institute (CLSI)*. <https://clsi.org/standards/products/microbiology/documents/m39/>.
|
|
#' @inheritSection AMR Read more on our website!
|
|
#' @examples
|
|
#' \donttest{
|
|
#' x <- bug_drug_combinations(example_isolates)
|
|
#' x
|
|
#' format(x, translate_ab = "name (atc)")
|
|
#'
|
|
#' # Use FUN to change to transformation of microorganism codes
|
|
#' x <- bug_drug_combinations(example_isolates,
|
|
#' FUN = mo_gramstain)
|
|
#'
|
|
#' x <- bug_drug_combinations(example_isolates,
|
|
#' FUN = function(x) ifelse(x == as.mo("E. coli"),
|
|
#' "E. coli",
|
|
#' "Others"))
|
|
#' }
|
|
bug_drug_combinations <- function(x,
|
|
col_mo = NULL,
|
|
FUN = mo_shortname,
|
|
...) {
|
|
stop_ifnot(is.data.frame(x), "`x` must be a data frame")
|
|
stop_ifnot(any(sapply(x, is.rsi), na.rm = TRUE), "No columns with class <rsi> found. See ?as.rsi.")
|
|
|
|
# try to find columns based on type
|
|
# -- mo
|
|
if (is.null(col_mo)) {
|
|
col_mo <- search_type_in_df(x = x, type = "mo")
|
|
}
|
|
stop_if(is.null(col_mo), "`col_mo` must be set")
|
|
|
|
x_class <- class(x)
|
|
x <- as.data.frame(x, stringsAsFactors = FALSE)
|
|
x[, col_mo] <- FUN(x[, col_mo, drop = TRUE], ...)
|
|
x <- x[, c(col_mo, names(which(sapply(x, is.rsi)))), drop = FALSE]
|
|
|
|
unique_mo <- sort(unique(x[, col_mo, drop = TRUE]))
|
|
|
|
out <- data.frame(
|
|
mo = character(0),
|
|
ab = character(0),
|
|
S = integer(0),
|
|
I = integer(0),
|
|
R = integer(0),
|
|
total = integer(0))
|
|
|
|
for (i in seq_len(length(unique_mo))) {
|
|
# filter on MO group and only select R/SI columns
|
|
x_mo_filter <- x[which(x[, col_mo, drop = TRUE] == unique_mo[i]), names(which(sapply(x, is.rsi))), drop = FALSE]
|
|
# turn and merge everything
|
|
pivot <- lapply(x_mo_filter, function(x) {
|
|
m <- as.matrix(table(x))
|
|
data.frame(S = m["S", ], I = m["I", ], R = m["R", ], stringsAsFactors = FALSE)
|
|
})
|
|
merged <- do.call(rbind, pivot)
|
|
out_group <- data.frame(mo = unique_mo[i],
|
|
ab = rownames(merged),
|
|
S = merged$S,
|
|
I = merged$I,
|
|
R = merged$R,
|
|
total = merged$S + merged$I + merged$R)
|
|
out <- rbind(out, out_group)
|
|
}
|
|
|
|
structure(.Data = out, class = c("bug_drug_combinations", x_class))
|
|
}
|
|
|
|
#' @method format bug_drug_combinations
|
|
#' @export
|
|
#' @rdname bug_drug_combinations
|
|
format.bug_drug_combinations <- function(x,
|
|
translate_ab = "name (ab, atc)",
|
|
language = get_locale(),
|
|
minimum = 30,
|
|
combine_SI = TRUE,
|
|
combine_IR = FALSE,
|
|
add_ab_group = TRUE,
|
|
remove_intrinsic_resistant = FALSE,
|
|
decimal.mark = getOption("OutDec"),
|
|
big.mark = ifelse(decimal.mark == ",", ".", ","),
|
|
...) {
|
|
x <- as.data.frame(x, stringsAsFactors = FALSE)
|
|
x <- subset(x, total >= minimum)
|
|
|
|
if (remove_intrinsic_resistant == TRUE) {
|
|
x <- subset(x, R != total)
|
|
}
|
|
if (combine_SI == TRUE | combine_IR == FALSE) {
|
|
x$isolates <- x$R
|
|
} else {
|
|
x$isolates <- x$R + x$I
|
|
}
|
|
|
|
give_ab_name <- function(ab, format, language) {
|
|
format <- tolower(format)
|
|
ab_txt <- rep(format, length(ab))
|
|
for (i in seq_len(length(ab_txt))) {
|
|
ab_txt[i] <- gsub("ab", as.character(as.ab(ab[i])), ab_txt[i])
|
|
ab_txt[i] <- gsub("cid", ab_cid(ab[i]), ab_txt[i])
|
|
ab_txt[i] <- gsub("group", ab_group(ab[i], language = language), ab_txt[i])
|
|
ab_txt[i] <- gsub("atc_group1", ab_atc_group1(ab[i], language = language), ab_txt[i])
|
|
ab_txt[i] <- gsub("atc_group2", ab_atc_group2(ab[i], language = language), ab_txt[i])
|
|
ab_txt[i] <- gsub("atc", ab_atc(ab[i]), ab_txt[i])
|
|
ab_txt[i] <- gsub("name", ab_name(ab[i], language = language), ab_txt[i])
|
|
ab_txt[i]
|
|
}
|
|
ab_txt
|
|
}
|
|
|
|
remove_NAs <- function(.data) {
|
|
cols <- colnames(.data)
|
|
.data <- as.data.frame(sapply(.data, function(x) ifelse(is.na(x), "", x), simplify = FALSE))
|
|
colnames(.data) <- cols
|
|
.data
|
|
}
|
|
|
|
create_var <- function(.data, ...) {
|
|
dots <- list(...)
|
|
for (i in seq_len(length(dots))) {
|
|
.data[, names(dots)[i]] <- dots[[i]]
|
|
}
|
|
.data
|
|
}
|
|
|
|
y <- x %>%
|
|
create_var(ab = as.ab(x$ab),
|
|
ab_txt = give_ab_name(ab = x$ab, format = translate_ab, language = language)) %>%
|
|
group_by(ab, ab_txt, mo) %>%
|
|
summarise(isolates = sum(isolates, na.rm = TRUE),
|
|
total = sum(total, na.rm = TRUE)) %>%
|
|
ungroup()
|
|
|
|
y <- y %>%
|
|
create_var(txt = paste0(percentage(y$isolates / y$total, decimal.mark = decimal.mark, big.mark = big.mark),
|
|
" (", trimws(format(y$isolates, big.mark = big.mark)), "/",
|
|
trimws(format(y$total, big.mark = big.mark)), ")")) %>%
|
|
select(ab, ab_txt, mo, txt) %>%
|
|
arrange(mo)
|
|
|
|
# replace tidyr::pivot_wider() from here
|
|
for (i in unique(y$mo)) {
|
|
mo_group <- y[which(y$mo == i), c("ab", "txt")]
|
|
colnames(mo_group) <- c("ab", i)
|
|
rownames(mo_group) <- NULL
|
|
y <- y %>%
|
|
left_join(mo_group, by = "ab")
|
|
}
|
|
y <- y %>%
|
|
distinct(ab, .keep_all = TRUE) %>%
|
|
select(-mo, -txt) %>%
|
|
# replace tidyr::pivot_wider() until here
|
|
remove_NAs()
|
|
|
|
select_ab_vars <- function(.data) {
|
|
.data[, c("ab_group", "ab_txt", colnames(.data)[!colnames(.data) %in% c("ab_group", "ab_txt", "ab")])]
|
|
}
|
|
|
|
y <- y %>%
|
|
create_var(ab_group = ab_group(y$ab, language = language)) %>%
|
|
select_ab_vars() %>%
|
|
arrange(ab_group, ab_txt)
|
|
y <- y %>%
|
|
create_var(ab_group = ifelse(y$ab_group != lag(y$ab_group) | is.na(lag(y$ab_group)), y$ab_group, ""))
|
|
|
|
if (add_ab_group == FALSE) {
|
|
y <- y %>%
|
|
select(-ab_group) %>%
|
|
rename("Drug" = ab_txt)
|
|
colnames(y)[1] <- translate_AMR(colnames(y)[1], language = get_locale(), only_unknown = FALSE)
|
|
} else {
|
|
y <- y %>% rename("Group" = ab_group,
|
|
"Drug" = ab_txt)
|
|
colnames(y)[1:2] <- translate_AMR(colnames(y)[1:2], language = get_locale(), only_unknown = FALSE)
|
|
}
|
|
|
|
rownames(y) <- NULL
|
|
y
|
|
}
|
|
|
|
#' @method print bug_drug_combinations
|
|
#' @export
|
|
print.bug_drug_combinations <- function(x, ...) {
|
|
x_class <- class(x)
|
|
print(structure(x, class = x_class[x_class != "bug_drug_combinations"]),
|
|
...)
|
|
message(font_blue("NOTE: Use 'format()' on this result to get a publishable/printable format."))
|
|
}
|